Beyond Distance and Speed Limit: Estimating Emergency Vehicle Travel Time Using Nationwide Data from Sweden

Abstract

Accurate travel time estimation for emergency vehicles is essential for effective dispatching and response planning. This paper investigates to what extent predictive performance improves when models incorporate variables beyond common inputs such as distance and speed limits. Using nationwide historical GPS data from Swedish emergency vehicles, map-matched trajectories were enriched with multiple variables, including speed limits, speed bumps, precipitation, and vehicle size. An automatic route validation model was developed to identify and remove flawed trajectories. The results show that extending models with additional variables improves estimation accuracy. A compact decision tree was among the best-performing models, improving a baseline model that uses only speed limits as input by 5.61 percentage points in MAPE and 20.26 seconds in RMSE, while more complex models offered limited additional benefits. These improvements reduce uncertainty in dispatch decisions and support more reliable route selection in time-critical emergency response operations.